Uniformize kwargs for processors - GroundingDINO

[SangbumChoi]

What does this PR do?

Fixes one model from #31911

We don;t want our tracker issue to be closed after the PR is merged, so I edited the magic keyword a bit

Before submitting

• [] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

@zucchini-nlp @NielsRogge @EduardoPach

[zucchini-nlp]

Thanks for helping us to make the processors more uniform!

Can you also add a test for this, it is available in:

from ...test_processing_common import ProcessorTesterMixin

[SangbumChoi]

@zucchini-nlp

transformers/tests/test_processing_common.py

Line 80 in 2c8180a

if hasattr(self, "tmpdirname"):

I think the circleCI fails since the upper PR is not merged in to main. Otherwise requesting for second review!

[SangbumChoi]

@zucchini-nlp Hi now I understand the full logic and had a one question on this PR.
It seems that initial PR was designed in here

https://github.com/huggingface/transformers/pull/31197/files#diff-767679f9407c55e91d247ac9ab5359a867414676ae6083959bd2c5cf4e1113c7

with based on following image processor

https://github.com/molbap/transformers/blob/bcce00766b8faa0bd48e0dfe4d0ec72db761f56a/src/transformers/models/efficientnet/image_processing_efficientnet.py

However, not all the processor has the crop_size argument for image_processor some of them has only size. I think in order to make it model-wise compatibility I recommend to change this to size.

https://github.com/molbap/transformers/blob/bcce00766b8faa0bd48e0dfe4d0ec72db761f56a/tests/test_processing_common.py

otherwise I can override these test for groundingdino only

WDYT? cc. @molbap

(BTW, I reverted all the irrelevant commit and major CI fail is related to this comment)

[zucchini-nlp]

IIRC the test will fail if we pass in both, crop_size and size, because one of the args will not be used by image processor. If that's the case, we can override since it is about making tests happy

[molbap]

Since only these two arguments are passed, we could also inspect in the test method and figure out which is used for a given model, then call with that one - skipping the test entirely is also acceptable, yes

[SangbumChoi]

IIRC the test will fail if we pass in both, crop_size and size, because one of the args will not be used by image processor. If that's the case, we can override since it is about making tests happy

Well this was not the exactly the case. As I mentioned GroundingDINO processor -> image_processor does not have function of crop_size. However, the major test in test_processor_common.py uses crop_size which is not used in the groundingdino. This will result the output shape as same as input since resizing technique is not applied.

I manually changed this to inherit size instead of crop_size to add resizing technique!

@zucchini-nlp Request for the review!

[zucchini-nlp]

Great work, thanks for working on it! Left one tiny comment

[molbap]

Thanks a lot @SangbumChoi 🙏
I added a comment as well on a test.

[zucchini-nlp]

Requesting review from core maintainer :)

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[amyeroberts]

Looks great - thanks for handling this!

[amyeroberts]

❤️